The Impact of Social Networks on the Stock Market Using Sentiment Analysis and Machine Learning: Application to the Turkish Stock Market

Abstract

The proliferation of portable devices and social media has transformed opinion sharing, impacting individual behavior, particularly in financial markets. This research explores how online sentiments influence investors' decision-making, highlighting the complexities of sentiment measurement in behavioral finance. AI-driven techniques have been developed to quantify opinions from social media data, focusing on Twitter (rebranded as X). The transformer architecture, which is a cutting-edge deep learning method widely used in generative AI models, is employed for sentiment analysis. The relationship between digitized sentiment scores and share prices within T & uuml;rkiye's Borsa & Idot;stanbul (BIST 30) index was analyzed using machine learning techniques. Social media activity, as indicated by tweet volume, was investigated in relation to stock prices. The dataset comprises nearly 1.9 million tweets related to BIST 30 stocks, collected from early 2021 to late 2022. Independent variables include tweet volume, sentiment (positivenegative), and tweet timing, whereas dependent variables comprise stock prices and index closures. The findings reveal that tweet volume effectively predicts stock prices. Positive sentiment demonstrates stronger predictive power for individual stocks, whereas overall tweet sentiment does not significantly affect index-wide prices. Conversely, tweet timing is ineffective for price prediction. This research exemplifies the growing application of AI and machine learning in the social sciences by quantifying human opinions. The proposed model offers both theoretical and practical contributions, serving as a model for future research while delivering new insights and recommendations. The insights gained underscore the potential to harness information systems to advance financial literacy, stimulate economic growth, and empower informed decision-making across diverse global contexts.